Domestic abuse is a pervasive global problem. Here we analyze two approaches to reducing violent DA recidivism. One involves charging the perpetrator with a crime; the other provides protective services to the victim on the basis of a formal risk assessment carried out by the police. We use detailed administrative data to estimate the average effect of treatment on the treated using inverse propensity-score weighting (IPW). We then make use of causal forests to study heterogeneity in the estimated treatment effects. We find that pressing charges substantially reduces the likelihood of violent recidivism. The analysis also reveals substantial heterogeneity in the effect of pressing charges. In contrast, the risk-assessment process has no discernible effect.